利用人工神经网络（ANN）技术，基于气象条件、污染物排放变化和污染物浓度资料构建污染物浓度统计模型，在此基础上分析气象条件和污染物排放源排放变化对污染物浓度逐日变化和年际变化的影响。研究结果发现基于ANN建立的大气污染统计预报模型模拟NO2浓度准确性较高，其次为SO2，PM10浓度准确性较低。ANN的输入参数更适合NO2的模拟，SO2和PM10浓度的影响因子较为复杂。气象条件变化是NO2浓度逐日变化的主要影响因子，污染物排放量变化是NO2浓度年际变化的主要影响因子。因子分离法计算得到的气象条件、污染物排放及两者相互作用对NO2浓度逐日变化的贡献率分别是57.9%、24.5%和17.6%，对NO2浓度年际变化的贡献率分别是13.7%、73.3%和13%。
To quantify the impact of meteorological conditions and pollutant emissions on air quality in Lanzhou, this paper developed an artificial neural network (ANN) model to forecast winter daily average pollutant concentrations in Lanzhou based on six years meteorology and pollutant concentration data, and the model was used to investigate the influence of meteorological conditions and pollutant emissions on daily and interannual variations of pollutant concentrations via sensitivity test. The high resolution meteorological data in Lanzhou was acquired from the Weather Research and Forecasting (WRF) model. The results showed that ANN model had a good performance to NO2, followed by SO2 and PM10. The statistical performance indicated that the input data selected in this study may be more suitable for NO2. The relative low statics for SO2 and PM10 were caused by the complex emissions for SO2 (elevated point sources) and PM10(local dust). With good performance, the NO2 was selected to analysis the influence of meteorological conditions and pollutant emissions. The change of meteorological conditions is the main factor causing the daily variation of NO2 concentration, while pollutant emissions change is mainly responsible for the interannual variation of NO2 concentration. Utilizing factor separation method, the contribution of meteorological conditions, pollutant emissions and interactions to NO2 concentration daily variation are 57.9%, 24.5% and 17.6%, respectively, and 13.7%, 73.3% and 13% for NO2 concentration interannual variation. The simple assumption of emission information has an adverse impact on the results, and the improvement of emission information will be needed in the further research.
[1]An X Q, Zuo H C, Chen L J. 2007. Atmospheric environmental capacity of SO<sub>2</sub> in winter over Lanzhou in China:A case study[J]. Adv Atmos Sci, 24(4):688-699.
[2]Andersson C, Langner J, Bergstrom R. 2007. Interannual variation and trends in air pollution over Europe due to climate variability during 19582001 simulated with a regional CTM coupled to the ERA40 reanalysis[J]. Tellus, 59B:77-98.
[3]Bandyopadhyay G, Chattopadhyay S. 2007. Single hidden layer artificial neural network models versus multiple linear regression model in forecasting the time series of total ozone[J]. Int J Environ Sci Technol, 4(1):141-149.
[4]Cai M, Yin Y F, Xie M. 2009. Prediction of hourly air pollutant concentrations near urban arterials using artificial neural network approach[J]. Transportation Research Part D, 14:32-41.
[5]Cermak J, Knutti R. 2009. Beijing Olympics as an aerosol field experiment[J]. Geophys Res Lett, 36:L10806.
[6]Chen Z H, Cheng S Y, Li J B, et al. 2008. Relationship between atmospheric pollution processes and synoptic pressure patterns in northern China[J]. Atmos Environ, 42:6078-6087.
[7]Cheng C S Q, Campbell M, Li Q, et al. 2007 A synoptic climatological approach to assess climatic impact on air quality in South-central Canada. Part I:Historical analysis[J]. Water, Air, and Soil pollution, 182(1-4):131-148.
[8]Fernando H J S, Mammarella M C, Grandoni G, et al. 2012. Forecasting PM<sub>10</sub> in metropolitan areas:Efficacy of neural networks[J]. Environmental Pollution, 163:62-67.
[9]Grivas G, Chaloulakou A. 2006. Artificial neural network models for prediction of PM<sub>10</sub> hourly concentrations, in the Greater Area of Athens, Greece[J]. Atmos Environ, 40:1216-1229.
[10]He J J, Yu Y, Liu N, et al. 2013. Numerical model based relationship between meteorological conditions and air quality and its implication for urban air quality management[J]. International Journal of Environment and Pollution, 53(3/4):265-286.
[11]Liu Y, He K, Li S, et al. 2012. A statistical model to evaluate the effectiveness of PM<sub>2.5</sub> emissions control during the Beijing 2008 Olympic Games[J]. Environmental International, 44:100-105.
[12]Nagendra S M S, Khare M. 2006. Artificial neural network approach for modeling nitrogen dioxide dispersion from vehicular exhaust emissions[J]. Ecological Modelling, 190:99-115.
[13]Panday A K, Prinn R G, Schar C. 2012. Diurnal cycle of air pollution in Kathmandu valley, Nepal:2. Modeling results[J]. J Geophys Res, 114:D21308.
[14]Pearce J L, Beringer J, Nicholls N, et al. 2011a. Investing the influence of synoptic-scale meteorology on air quality using self-organizing maps and generalized additive modelling[J]. Atmos Environ, 45:128-136.
[15]Pearce J L, Beringer J, Nicholls N, et al. 2011b. Quantifying the influence of local meteorology on air quality using generalized additive models[J]. Atmos Environ, 45:1328-1336.
[16]Roustan Y, Pausader M, Seigneur C. 2011. Estimating the effect of on-road vehicle emission controls on future air quality in Paris, France[J]. Atmos Environ, 45:6828-6836.
[17]Stein U, Alpert P. 1993. Factor separation in numerical simulations[J]. Journal of the Atmospheric Sciences, 50(14):2107-2115.
[18]Wang W, Primbs T, Tao S, et al. 2009. Atmospheric particulate matter pollution during the 2008 Beijing Olympics[J]. Environ Sci Technol, 43(14):5314-5320.
[19]Wang X M, Chen F, Wu ZY, et al. 2009. Impacts of Weather Conditions Modified by Urban Expansion on Surface Ozone:Comparsion between the Pearl River Delta and Yangtze River Delta Regions[J]. Adv Atmos Sci, 26(5):962-972.
[20]Zhang J P, Zhu T, Zhang Q H, et al. 2012. The impact of circulation patterns on regional transport pathways and air quality over Beijing and its surrounding[J]. Atmospheric Chemistry and Physics, 12:5031-5051.
[21]Bai Xiaoping, Li Hong, Zhang Qiming, et al. 2006. Progress of research on artificial neural network in air pollution prediction[J]. Science & Technology Review, 24(12):77-81.<br/>白晓平, 李红, 张启明, 等. 2006. 人工神经网络在空气污染预报中的研究进展[J]. 科技导报, 24(12):77-81.
[22]Chen Leihua, Yu Ye, Chen Jinbei, et al. 2010. Characteristics of main air pollution in Lanzhou during 2001-2007[J]. Plateau Meteor, 29(06):1627-1633.<br/>陈雷华, 余晔, 陈晋北, 等. 2010.2001-2007年兰州市主要大气污染物污染特征分析[J]. 高原气象, 29(06):1627-1633.
[23]Guo Qingchun, He Zhenfang, Kou Liqun, et al. 2011. Application of BP neural network to forecasting API in Beijing[J]. Environmental Engineering, 29(4):106-108.<br/>郭庆春, 何振芳, 寇立群, 等. 2011. BP 神经网络在北京市API预报中的应用[J]. 环境工程, 29(4):106-108.
[24]Hu Yinqiao, Zhang Qiang. 1999. Atmosphere pollution mechanism along with prevention and cure countermeasure of the Lanzhou hollow basin[J]. China Environmental Science, 19(2):119-122.<br/>胡隐樵, 张强. 1999. 兰州市大气污染的物理机制与防治对策[J]. 中国环境科学, 19(2):119-122.
[25]Xia Zhiye, Liu Zhihong, Wang Yongqian, et al. 2015. Research on ground-level PM<sub>2.5</sub> mass concentration retrieval based on MODIS aerosol optical thickness[J]. Plateau Meteor, 34(6):1765-1771. DOI:10.7522/j. issn. 1000-0534.2014.00075.<br/>夏志业, 刘志红, 王永前, 等. 2015. MODIS气溶胶光学厚度的PM<sub>2.5</sub>质量浓度遥感反演研究[J]. 高原气象, 34(6):1765-1771.
[26]Zhang Yuanhang, Tang Xiaoyan, Bi Mutian, et al. 1987. Identification of pollutant sources of aerosol in Xigu area, Lanzhou[J]. Acta Scientia Circumstantiae, 7(3):269-278.<br/>张远航, 唐孝炎, 毕木天, 等. 1987. 兰州西固地区气溶胶污染源的鉴别[J]. 环境科学学报, 7(3):269-278.
